Heads up.... "Intel play down Alder Lake CPU bending"

[hardcorefs]

Looks like Intel might have cheap assed the design of the 12th gen

Intel Comments On Alder Lake's Warping and Bending Issues, Mods Void Warranty

"Nothing to see here, folks"

www.tomshardware.com

 

[Leesureone]

Looks like Intel might have cheap assed the design of the 12th gen

Intel Comments On Alder Lake's Warping and Bending Issues, Mods Void Warranty

"Nothing to see here, folks"

www.tomshardware.com

Hi @hardcorefs, that "bending" was reported quite a while ago but hasn't exactly turned into "Bendgate" if you will. The article link you posted is a couple of months old and was published because intel formally addressed what had been reported. There's always some spin on these kinds of issues but here within the community no users, including myself, have had anything but good performance using a variety of coolers.

 

[dehjomz]

My alder lake runs at 26-27 degrees at idle. 45 - 48 degrees during gaming. A Cinebench R23 all core load which I never do runs at approx 81 degrees max. Alder lake is cooler at idle than all the previous lakes I’ve owned, and runs cooler during gaming, and is roughly the same temp under an all core load.

I’d say “bendgate” isnt really a scandal at all, but perhaps the design of the loading mechanism could be optimized in some motherboards. There’s always room for improvement.

 

[craighazan]

I prefer "warp gate"!.

 

[Leesureone]

My alder lake runs at 26-27 degrees at idle. 45 - 48 degrees during gaming. A Cinebench R23 all core load which I never do runs at approx 81 degrees max. Alder lake is cooler at idle than all the previous lakes I’ve owned, and runs cooler during gaming, and is roughly the same temp under an all core load.

I’d say “bendgate” isnt really a scandal at all, but perhaps the design of the loading mechanism could be optimized in some motherboards. There’s always room for improvement.

It’s have almost identical thermals

 

[c-o-pr]

My alder lake runs at 26-27 degrees at idle. 45 - 48 degrees during gaming. A Cinebench R23 all core load which I never do runs at approx 81 degrees max.

Please, "81" at what clock rate using what cooling given what ambient?

Temps are meaningless without such context.

The concern about bending was a demonstration that pressure around the CPU package can warp it, causing poor alignment of the heat spreader to the lid, and by implication possible alignment issues within the package, causing overall poorer thermal conductivity and reduced max power. I recall seeing a picture of a gap caused by flexing, and using washers to reduce stress around the socket with a claim of a very few degrees C of reduction in cooling at high power.

What it comes down to is headroom at highest clock rates.

It is meaningless that the chip has a given idle temp, and almost meaningless that it has a given load temp under ordinary operating conditions. If temp is super high under non-stressful loads, this indicates a defect in the cooler or its attachment to the CPU, but this is only significant because the chip is clocked to a certain power.

This is the true meaning of "unlocked" which is permit exceeding rated power by taking more responsibility for thermals.

What matters is hitting top overclocking rate without busting stability or hitting the thermal rev limiter.

Bending (and de-lidding) matter at highest power, because these traits can make the diff between stable and unstable kit.

But this is just because there's little headroom on thermals at highest clocking, not because any temperature within norms is better than any other.

As there's a literal causality between work rate and temps, then anything that makes CPU work harder incurs a directly proportional thermal load: Make it work harder, temps go up.

There's a fixed surface area through which heat can be conducted given a perfect coupling of the cooler to the die which limits power (clock rate). So bendgate is a concern about a reduction in this limit due to poor coupling.

From here you can see that reporting XYZ temp alone offers nothing to understanding of the system.

To be helpful, report a temp at some clock rate for some workload so we can surmise as to whether peak performance is limited by temp. From there we can wonder about cooler type, bending, paste, voltages, lids, etc.

 

[UtterDisbelief]

Please, "81" at what clock rate using what cooling given what ambient?

Temps are meaningless without such context.

The concern about bending was a demonstration that pressure around the CPU package can warp it, causing poor alignment of the heat spreader to the lid, and by implication possible alignment issues within the package, causing overall poorer thermal conductivity and reduced max power. I recall seeing a picture of a gap caused by flexing, and using washers to reduce stress around the socket with a claim of a very few degrees C of reduction in cooling at high power.

What it comes down to is headroom at highest clock rates.

It is meaningless that the chip has a given idle temp, and almost meaningless that it has a given load temp under ordinary operating conditions. If temp is super high under non-stressful loads, this indicates a defect in the cooler or its attachment to the CPU, but this is only significant because the chip is clocked to a certain power.

This is the true meaning of "unlocked" which is permit exceeding rated power by taking more responsibility for thermals.

What matters is hitting top overclocking rate without busting stability or hitting the thermal rev limiter.

Bending (and de-lidding) matter at highest power, because these traits can make the diff between stable and unstable kit.

But this is just because there's little headroom on thermals at highest clocking, not because any temperature within norms is better than any other.

As there's a literal causality between work rate and temps, then anything that makes CPU work harder incurs a directly proportional thermal load: Make it work harder, temps go up.

There's a fixed surface area through which heat can be conducted given a perfect coupling of the cooler to the die which limits power (clock rate). So bendgate is a concern about a reduction in this limit due to poor coupling.

From here you can see that reporting XYZ temp alone offers nothing to understanding of the system.

To be helpful, report a temp at some clock rate for some workload so we can surmise as to whether peak performance is limited by temp. From there we can wonder about cooler type, bending, paste, voltages, lids, etc.

That's a great overview of the subject.

I would, however, think it's relevant if for example I tell you my CPU idles at 30-degrees-C and yet yours, which is in an identical build, runs at 25-degrees-C. That implies that the way I have put the components together is possibly not so good, or with as much care than yours. Hypothetically.

It can also help people choose a specific case or CPU cooler.

If the reason for the comparison is just bragging-rights, then yes, pointless.

 

[pastrychef]

My alder lake runs at 26-27 degrees at idle. 45 - 48 degrees during gaming. A Cinebench R23 all core load which I never do runs at approx 81 degrees max. Alder lake is cooler at idle than all the previous lakes I’ve owned, and runs cooler during gaming, and is roughly the same temp under an all core load.

I’d say “bendgate” isnt really a scandal at all, but perhaps the design of the loading mechanism could be optimized in some motherboards. There’s always room for improvement.

What kind of cooler are you using? Did you use the same cooler on your previous lakes?

 

[Leesureone]

Please, "81" at what clock rate using what cooling given what ambient?

Temps are meaningless without such context.

The concern about bending was a demonstration that pressure around the CPU package can warp it, causing poor alignment of the heat spreader to the lid, and by implication possible alignment issues within the package, causing overall poorer thermal conductivity and reduced max power. I recall seeing a picture of a gap caused by flexing, and using washers to reduce stress around the socket with a claim of a very few degrees C of reduction in cooling at high power.

What it comes down to is headroom at highest clock rates.

It is meaningless that the chip has a given idle temp, and almost meaningless that it has a given load temp under ordinary operating conditions. If temp is super high under non-stressful loads, this indicates a defect in the cooler or its attachment to the CPU, but this is only significant because the chip is clocked to a certain power.

This is the true meaning of "unlocked" which is permit exceeding rated power by taking more responsibility for thermals.

What matters is hitting top overclocking rate without busting stability or hitting the thermal rev limiter.

Bending (and de-lidding) matter at highest power, because these traits can make the diff between stable and unstable kit.

But this is just because there's little headroom on thermals at highest clocking, not because any temperature within norms is better than any other.

As there's a literal causality between work rate and temps, then anything that makes CPU work harder incurs a directly proportional thermal load: Make it work harder, temps go up.

There's a fixed surface area through which heat can be conducted given a perfect coupling of the cooler to the die which limits power (clock rate). So bendgate is a concern about a reduction in this limit due to poor coupling.

From here you can see that reporting XYZ temp alone offers nothing to understanding of the system.

To be helpful, report a temp at some clock rate for some workload so we can surmise as to whether peak performance is limited by temp. From there we can wonder about cooler type, bending, paste, voltages, lids, etc.

You are talking about a next level analysis of thermals which could be interesting. Personally I don't feel its necessary to prove to me that "Bendgate" isn't a huge issue. I've had the pleasure of building four different Z690 systems now and all thermals, using varying components, were very similar. All of then consistent with Intels published specs for the various processors. In disassembly of more than one build I've found consistent coverage, or contact, of the thermal paste on both surfaces (at least to the naked eye). My point being there was no evidence of warping nor did I experience cooling issues under varying loads. @CaseySJ expressed similar confidence in the three systems he's built saying it has not been an issue.

 

[dehjomz]

What kind of cooler are you using? Did you use the same cooler on your previous lakes?

I’m using a nzxt x63 for cooling alder lake, same as on Rocket lake and comet lake. Skylake was a different cooler.

Please, "81" at what clock rate using what cooling given what ambient?

Temps are meaningless without such context.

The concern about bending was a demonstration that pressure around the CPU package can warp it, causing poor alignment of the heat spreader to the lid, and by implication possible alignment issues within the package, causing overall poorer thermal conductivity and reduced max power. I recall seeing a picture of a gap caused by flexing, and using washers to reduce stress around the socket with a claim of a very few degrees C of reduction in cooling at high power.

What it comes down to is headroom at highest clock rates.

It is meaningless that the chip has a given idle temp, and almost meaningless that it has a given load temp under ordinary operating conditions. If temp is super high under non-stressful loads, this indicates a defect in the cooler or its attachment to the CPU, but this is only significant because the chip is clocked to a certain power.

This is the true meaning of "unlocked" which is permit exceeding rated power by taking more responsibility for thermals.

What matters is hitting top overclocking rate without busting stability or hitting the thermal rev limiter.

Bending (and de-lidding) matter at highest power, because these traits can make the diff between stable and unstable kit.

But this is just because there's little headroom on thermals at highest clocking, not because any temperature within norms is better than any other.

As there's a literal causality between work rate and temps, then anything that makes CPU work harder incurs a directly proportional thermal load: Make it work harder, temps go up.

There's a fixed surface area through which heat can be conducted given a perfect coupling of the cooler to the die which limits power (clock rate). So bendgate is a concern about a reduction in this limit due to poor coupling.

From here you can see that reporting XYZ temp alone offers nothing to understanding of the system.

To be helpful, report a temp at some clock rate for some workload so we can surmise as to whether peak performance is limited by temp. From there we can wonder about cooler type, bending, paste, voltages, lids, etc.

Cinebench r23 is the same test I’ve ran on z490, z590, and now z690. For alder lake, I restricted the all-core p-core frequency to 5.0 GHz and the e-core frequency to 3.9. I’ve successfully ran the test with cores at 5.1 and e-cores at 4.0, but that adds additional heat for barely any more performance. Same cooler brand (nzxt x63), same thermal paste from thermal grizzly. Same fans and fan speeds. For alder lake, I ordered an asetek lga1700 mounting mechanism for the cooler. That probably helped with temps.

At idle in windows and Linux, the schedulers appear to schedule threads on the E cores, and clocks get as low as 800 MHz on the P cores and 400 MHz on the E cores that could be part of why Alder Lake idles cooler. But even in macOS with no kernel understanding of an x86 e-core, the system idles cooler at 27-28 degrees.

The simple point I’m making is alder lake runs no hotter than previous lakes I’ve owned, in my experience, and even idles cooler. Same room. Same case. Same ambient temps although now is summer so it’s hotter.

I should add that I purchased i7s. Even though I can afford i9s I usually skip them, cause the i7 is the best bang for buck in my opinion. Almost if not the same i9 performance at a lower cost and i9s probably run hotter under load.

No bragging, just my lived experience. If bend gate were a universal problem, I’d expect my temps to be hotter than before but they’re not. But on the other hand, perhaps temps are hotter than they otherwise need to be.